A common format to encode 360 degree virtual reality (VR) content into a classic, rectangular two dimensional (2D) image is using an “equirectangular” format. Similar to classic map projections used in cartography, the equirectangular format aims to represent data from the surface of a sphere as accurately as possible by giving all areas of the sphere a similar area on the rectangular surface of the 2D image.
The resulting 2D image features the original content with severely distorted straight lines, which may be similar to the distorted lines in an extreme wide angle photo. The distorted lines may be decoded into an un-distorted three dimensional (3D) image at viewing time; however, the distorted lines present particular difficulties when encoding the image as part of a video. For example, the distorted nature of the image leads to increased data being needed to represent the video in encoded format at a given quality. The additional data required increases the data that needs to be encoded, which uses more bandwidth making it a challenge to offer the video for the virtual reality experience at an acceptable bandwidth for streaming. Further, when the virtual reality experience is stereoscopic (e.g., individual images for each eye), the problem is exacerbated because image data for each eye needs to be encoded and transmitted.